Using Python3 To Estimate BIS Gear Sets For Ninja.
Langues: JP EN DE FR
users online
Forum » FFXI » Jobs » Ninja » Using Python3 to estimate BIS gear sets for Ninja.
Using Python3 to estimate BIS gear sets for Ninja.
First Page 2 3 4 5 6 7 8 9
 Cerberus.Shadowmeld
Offline
Serveur: Cerberus
Game: FFXI
Posts: 1667
By Cerberus.Shadowmeld 2023-01-04 11:33:04
Link | Citer | R
  
I was playing around with this, and it's got some very odd results. I was trying to see what some decent options were for Shining One Impulse drive, and some custom "hypothetical" stats.

Ended up with higher maximum/minimum (same set) when I untick blaze of glory fury, vs. having it ticked. Shows the same attack value for both in the distrobution graphs.

Can't remember how to store pics... lol

https://imgur.com/a/i5bEqGr
 Izanami
Offline
Posts: 298
By Izanami 2023-01-04 11:49:08
Link | Citer | R
  
Since this is a new page, I'll use this post to remind users that my most recent update to the code was posted 2023 January 03. The details of the update can be found on the previous page here.



Cerberus.Shadowmeld said: »
I was playing around with this, and it's got some very odd results. I was trying to see what some decent options were for Shining One Impulse drive, and some custom "hypothetical" stats.

Ended up with higher maximum/minimum (same set) when I untick blaze of glory fury, vs. having it ticked. Shows the same attack value for both in the distrobution graphs.

It looks like you're only running 100 simulations. You'll need to run 10,000 to get "decently accurate" statistics. I recommend running 50,000 simulations, though.
Setting 50,000 simulations for the output distribution.

I'll look into what BoG is (or isn't) doing in a bit.

Edit: Blaze of Glory is working as intended. Your Geo-Frailty bubble is getting BoG (+50% potency), but the Indi-Fury is not. BoG only applies to Geo-Spells in game, so I have it only apply to the Geo-Spell in the code too.

I do similar things for Marcato and Crooked Cards. Only the first roll/song get these bonuses.
 Cerberus.Shadowmeld
Offline
Serveur: Cerberus
Game: FFXI
Posts: 1667
By Cerberus.Shadowmeld 2023-01-04 12:40:00
Link | Citer | R
  
Oh, that was a dumb on my part, I had removed the geo bubble entirely, and just had indi. I meant to make it geo-fury
 zixxer
Offline
Posts: 249
By zixxer 2023-01-04 14:33:11
Link | Citer | R
  
Love the program! Would you be able to do this same thing for optimizing TP?

Happy new year!
 Cerberus.Shadowmeld
Offline
Serveur: Cerberus
Game: FFXI
Posts: 1667
By Cerberus.Shadowmeld 2023-01-04 15:19:21
Link | Citer | R
  
Has anyone done any significant work to get a good idea of what Odyssey Boss stats are at. I just put in random of 2000 defense ~350 VIT but the numbers this is giving me still seems really high.
 Izanami
Offline
Posts: 298
By Izanami 2023-01-04 23:54:58
Link | Citer | R
  
I've fixed a bug in the code that caused Shining One's Crit Rate to not take effect. The code was technically calculating the crit rate, but it was not applying it to any weapon skills. The code on the GitHub page has been updated.

See below for the "best" (R30 Odyssey) sets for Warrior Impulse Drive after applying the fix for Shining One. I do not play Warrior, so let me know if there is anything else that feels "off" about these sets and I'll look into potential issues related to it. I should note that I only looked into a potential issue with Shining One because somebody noticed that the Empyrean +3 feet (crit damage +13%) were not showing up in the original sets.


Impulse Drive: 2000-2300 TP
ItemSet 389196 ItemSet 389195

• Cichol: STR+Crit
• R25: Hjarrandi breastplate
• Cichol: STR+WSD



There hasn't been time to do much else to the code since I just posted the bigger update yesterday, but I've made the following changes in addition to the fix for Shining One.

  • Subtle Blow and Subtle Blow II have been added to relevant gear in preparation for TP sets.

  • I've added checkboxes for WAR's Blood Rage (+20% Crit Rate) and Mighty Strikes (+100% Crit Rate)

    • Mighty Strikes grants 100% crit rate to all weapon skills, even those that can not natively crit.


As always: keep letting me know if you come across any issues. These could be typos in gear, gearsets that don't feel right, or general errors the code shouldn't be throwing at you.



zixxer said: »
Love the program! Would you be able to do this same thing for optimizing TP?
Yes. I'm working on this now, but I'm currently stuck on how to efficiently find the highest damage set that also has, for example, -50 DT and +50 Subtle Blow.

If anyone has ideas for how to programmatically handle this, let me know. As far as I can tell, there are too many combinations of gear that would satisfy any defensive criteria to efficiently check each while maximizing damage dealt.

I'll likely first post the basic TP set finder and let the user manually make adjustments for things like DT, Magic Evasion, and Subtle Blow. The more advanced TP set finder will take much longer to create unless I can quickly solve the current issue.


Cerberus.Shadowmeld said: »
Has anyone done any significant work to get a good idea of what Odyssey Boss stats are at. I just put in random of 2000 defense ~350 VIT but the numbers this is giving me still seems really high.

I haven't found a source for Defense and VIT for Odyssey Gaol bosses. Hopefully somebody has done some testing and can provide rough estimates. I imagine defense would be at least 4000~6000 for v25s.
[+]
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-05 01:38:17
Link | Citer | R
  
Izanami said: »
If anyone has ideas for how to programmatically handle this, let me know.
The best way I can think of would be a full re-write of most of the code, since your program swaps items one a time to find upgrades.
 SimonSes
Offline
Posts: 8846
By SimonSes 2023-01-05 03:51:03
Link | Citer | R
  
Izanami said: »
I haven't found a source for Defense and VIT for Odyssey Gaol bosses. Hopefully somebody has done some testing and can provide rough estimates. I imagine defense would be at least 4000~6000 for v25s.

I think they just have some DT and it's not a problem of cRatio. Someone needs to go with BRD RNG GEO DNC RDM COR to maybe Arebati and check if you cap attack on RNG (same DMG with each shot) with bolster frailty, dia III and box step and all the attack buffs possible. If attack will be capped, check how much damage you do and compere to how much you should be doing.
 Zannakun
Offline
Posts: 6
By Zannakun 2023-01-05 07:47:36
Link | Citer | R
  
Izanami said: »
If anyone has ideas for how to programmatically handle this, let me know. As far as I can tell, there are too many combinations of gear that would satisfy any defensive criteria to efficiently check each while maximizing damage dealt.

I think you'd have to brute force the calculations. If that ends up being the case, then I'd go with multi-threading processing of sets.

Also, would need to have a point priority system. For example, let's say each point of DT prior to 50% is worth 10 points (configurable), Store TP worth (25 *((4/HitsToWS) + 1) ) points, and damage dealt worth (DMG Dealt / 1000) points.
 Izanami
Offline
Posts: 298
By Izanami 2023-01-05 09:21:53
Link | Citer | R
  
Zannakun said: »
I think you'd have to brute force the calculations. If that ends up being the case, then I'd go with multi-threading processing of sets.

I typically jump straight into the "brute force" method as a first-step for programming, then move to efficiency after I have a working version of whatever I'm working on. However, in this case, I've noticed that brute force will not be feasible. GPU acceleration at a high-performance computing center might work (something like 200,000+ cores available per partition), but the standard 2~32 CPU cores that most users would have access to will not be nearly sufficient.

My first idea was to create a list of all combinations of gear that could satisfy some criteria, such as DT<-40 and Subtle Blow>50. I would run the main code for each of these sets and return the one set with the highest damage (or lowest time to TP). This method would definitely work for finding the "best" set given some conditions. The number of sets changes per job since each job has access to different gear, but taking MNK for example:

If I require that the best set ONLY has at least "25 Subtle Blow" and "25 Subtle Blow II", then I'm seeing 142,848 valid gear sets that satisfy the criteria. This is ignoring all slots that do not have any Subtle Blow or Subtle Blow II options. I would then need to run the main code (while ignoring the slots with Subtle Blow) for each of these sets to find the one with the highest damage.

If I also wanted the set to have -50% PDT and -20% MDT, then the number of valid combinations is too high for my computer to find within the 10 minutes before I killed the process. The is because the code must first create all possible combinations of gear with SB, SB2, PDT, MDT, PDT2, and DT, then filter it based on the criteria to find the valid sets.


Zannakun said: »
Also, would need to have a point priority system. For example, let's say each point of DT prior to 50% is worth 10 points (configurable), Store TP worth (25 *((4/HitsToWS) + 1) ) points, and damage dealt worth (DMG Dealt / 1000) points.

I feel like a point-based priority/weight system would break down with varying buffs. It would also not guarantee that the final set satisfies whatever criteria the user provides, given that such a set exists in the first place.



Ramuh.Austar said: »
The best way I can think of would be a full re-write of most of the code, since your program swaps items one a time to find upgrades.

What did you have in mind? The automated part of the code is just a triple for-loop for swapping up to three pieces of equipment simultaneously. The end of the loop simply builds the set and runs an average damage test on it. It only looks complicated because I included so many conditions to remove invalid sets (two Gere Rings equipped at the same time, for example). This is probably the only part of the code that would need to be rewritten and it wouldn't be a problem at all.



My current idea is to run the main code once to find the best set ignoring the user's conditions, then rerun the code multiple times more while slowly increasing the requirements on the conditions.

For example, if we wanted to find the best set with 50 Subtle Blow:

  1. Find the best set with no conditions. Record how much Subtle Blow it already has.

  2. Reset best set damage (because it will certainly be decreasing with defensive swaps).

  3. Require that the new best set has at least "+X" more Subtle Blow than the original/previous best set, rejecting all sets that do not.

  4. Repeat steps 2 and 3 until the best set satisfies the user's conditions.

  5. Break out of the loop if there are no more available Subtle Blow swaps.



This method requires that X be small enough that 1~2 pieces of gear together can cover the difference (otherwise the code won't see gear with smaller bonuses as an option), but large enough that we will not have to run the code 10+ times to build up to the required conditions. If the first best set has 0 Subtle Blow (no traits either), then if we simply use X=5, then the code may have to run up to 10 times to build up to +50 Subtle Blow.

This would likely increase the code's runtime from ~30s per set to maybe 3~5 min per set, which is annoying, but workable. The runtime would not necessarily increase 10x if the code needed to run 10 times since the code would reject most sets for failing the conditional check before testing them. This would also only involve adding a fourth for-loop and an if-statement to reject sets that do not satisfy the criteria.
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-05 10:06:12
Link | Citer | R
  
Izanami said: »
What did you have in mind?
well, since you're just going to use the average in every step like the spreadsheet would, you'd make functions to get all of those averages together. and once you have that, you can set it up in a way similar to my set optimizer programs with conditions using list comprehension and lambda. it already works for stuff like finding spell sets for tanks that require x HP, x PDT, and maximizing fast cast stat from there, only you'd just change the maximize fast cast stat to maximize damage.
 Asura.Cluste
Offline
Serveur: Asura
Game: FFXI
user: Bjggi
Posts: 15
By Asura.Cluste 2023-01-05 12:51:11
Link | Citer | R
  
Could Tachi: Kagero be added to the List?

Playing with some more SAM Sets, what i find interesting. With Magic Buffs and Tachi: Jinpu. The Fotia Gorget (by alot) and Acc/Att/WSD Back is winning over Neck+2 and Macc/Mad/WSD Back.

Guess i have to make now a High/Mid Buffed Jinpu Set :D

Bc when you change some of the Magic Buffs to Attack Buffs it changes to Neck+2 and MAD Back.

Thanks again for your work! Looking forward to the TP Section.
 Izanami
Offline
Posts: 298
By Izanami 2023-01-05 22:25:55
Link | Citer | R
  
Ramuh.Austar said: »
Izanami said: »
What did you have in mind?
well, since you're just going to use the average in every step like the spreadsheet would, you'd make functions to get all of those averages together. and once you have that, you can set it up in a way similar to my set optimizer programs with conditions using list comprehension and lambda. it already works for stuff like finding spell sets for tanks that require x HP, x PDT, and maximizing fast cast stat from there, only you'd just change the maximize fast cast stat to maximize damage.
I must not understand what you're saying because it sounds like you described the brute force method I mentioned before. If I'm going to create a list of average damage for each valid set just to find the best, then the first step is to create all possible valid sets that satisfy the conditions. How many sets does your optimizer consider? I'm seeing an unreasonable amount for Subtle Blow alone.

For MNK wanting only "Subtle Blow >= 50" and "Subtle Blow II >= 25", I'm seeing 1,548,288 combinations of gear with subtle blow or subtle blow II and 141,096 valid sets that satisfy the conditions (ignoring slots without any subtle blow options). Each of these valid sets needs to be run through the automated set finder to fill in the non-subtle blow slots and maximize damage. Even assuming that each valid set only takes 1 second to find the best damage through the automated set finder, it'll still take 141,096/86,400 = 1.6 days to find the one best set.

The total number of sets grows exponentially as I add more stats to check. If I wanted all combinations of Subtle Blow/II and PDT/MDT/DT, then the code finds 1.3 trillion combinations, and even waiting 3.5 hours while I was away from home wasn't enough to find the number of sets that satisfied the conditions; it was still checking somewhere between the 9th and 10th billionth set. It probably would've taken another ~19 days to just find the valid sets from these.


Asura.Cluste said: »
Could Tachi: Kagero be added to the List?

It looks like I added Kagero a long time ago, but never made it selectable in the drop down list. I've added the following weapon skills to the code:

  • Katana: Retsu, Jin, Teki, To

  • Great Katana: Koki, Goten, Kagero, Enpi, Yukikaze

  • Dagger: Viper Bite

  • Sword: Circle Blade, Swift Blade, Fast Blade, Burning Blade, Fast Blade II

  • Scythe: Slice, Spinning Scythe

  • Great Sword: Shockwave, Sickle Moon, Spinning Slash, Hard Slash

  • Club: Shining Strike, True Strike, Mystic Boon

  • Polearm: Double Thrust

  • Staff: Full Swing, Sunburst, Starburst, Heavy Swing, Gate of Tartarus, Rock Crusher

  • Great Axe: Steel Cyclone, Iron Tempest

  • Axe: Spinning Axe, Raging Axe

  • Archery: Sidewinder, Blast Arrow, Piercing Arrow

  • Marksmanship: Blast Shot, Slug Shot, Split Shot

  • Hand-to-Hand: Spinning Attack, Combo, One Inch Punch


I've also added a few of the su2 weapons which grant access to some of these weapon skills.

Available weapon skills (not sorted)
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-05 22:33:31
Link | Citer | R
  
Izanami said: »
For MNK wanting only "Subtle Blow >= 50" and "Subtle Blow II >= 25", I'm seeing 1,548,288 combinations of gear with subtle blow or subtle blow II and 141,096 valid sets that satisfy the conditions (ignoring slots without any subtle blow options). Each of these valid sets needs to be run through the automated set finder to fill in the non-subtle blow slots and maximize damage. Even assuming that each valid set only takes 1 second to find the best damage through the automated set finder, it'll still take 141,096/86,400 = 1.6 days to find the one best set.
My old *** 2011 PC just ran 144018 sets in less than a second from one of my optimizers, and I only print the 'valid' total, that's not counting every combination that it checks to find the 'valid' amounts.
[+]
 Izanami
Offline
Posts: 298
By Izanami 2023-01-06 00:21:25
Link | Citer | R
  
Ramuh.Austar said: »
Izanami said: »
For MNK wanting only "Subtle Blow >= 50" and "Subtle Blow II >= 25", I'm seeing 1,548,288 combinations of gear with subtle blow or subtle blow II and 141,096 valid sets that satisfy the conditions (ignoring slots without any subtle blow options). Each of these valid sets needs to be run through the automated set finder to fill in the non-subtle blow slots and maximize damage. Even assuming that each valid set only takes 1 second to find the best damage through the automated set finder, it'll still take 141,096/86,400 = 1.6 days to find the one best set.
My old *** 2011 PC just ran 144018 sets in less than a second from one of my optimizers, and I only print the 'valid' total, that's not counting every combination that it checks to find the 'valid' amounts.

Unless I'm missing some incredibly useful Python tools, then I think we're talking about different things. I'm hoping that I'm just missing something due to my inexperience with Python, because whatever you're doing sounds incredibly useful.

If I use itertools to build the combinations of all possible sets with some level of subtle blow, dt, pdt, and mdt, then the code starts storing sets on disk because my 16GB of RAM is not enough to store 1.3 trillion sets worth of data. I'm not sure how much memory it would take to store 1.3 trillion gear set dictionaries, but numpy.zeros() wants 9.4 TiB for 1.3 trillion float64 zeros, or 4.7 TiB for int32 zeros. I can get around this memory issue by running a for loop, but it would take about 19 days to finish.

Code 
numpy.core._exceptions.MemoryError: Unable to allocate 9.46 TiB for an array with shape (1300000000000,) and data type float64
numpy.core._exceptions.MemoryError: Unable to allocate 4.73 TiB for an array with shape (1300000000000,) and data type int32

Finding a "best" Blade: Ten weapon skill set, for example, typically takes the GUI/code 20~30 seconds on my computer when checking all 16 slots. If I lock half of these slots to account for the fixed part of the set that stores Subtle Blow (and/or DT, Meva, etc), then the runtime drops to 4~6 seconds. That's still ~8 days to find the best set for each combination of Subtle Blow gear that satisfies the conditions. It's only after these 8 days that I could find the true best set out of the 140,000 possibilities.

I have a hard time believing that your computer can find all 140,000 best sets and return the true best set in less than one second. What method are you using exactly to even create the 1 trillion possible combinations so quickly?
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-06 01:02:52
Link | Citer | R
  
Izanami said: »
Unless I'm missing some incredibly useful Python tools, then I think we're talking about different things. I'm hoping that I'm just missing something due to my inexperience with Python, because whatever you're doing sounds incredibly useful.
I just use list comprehension and lambda.

I set a condition in my Snapshot optimizer, turned on all pieces of gear and set the condition to >= Snapshot, so it would find every combination. It ran 544320 sets in 1.74 seconds. And my computer has 4G of RAM for reference, it's ***.

Ideally, you'd be able to turn off pieces of gear you have no use for or are redundant, you wouldn't need Rajas Ring on MNK to cap Subtle Blow, for example, so don't consider it.
 Izanami
Offline
Posts: 298
By Izanami 2023-01-06 08:55:38
Link | Citer | R
  
Ramuh.Austar said: »
Izanami said: »
Unless I'm missing some incredibly useful Python tools, then I think we're talking about different things. I'm hoping that I'm just missing something due to my inexperience with Python, because whatever you're doing sounds incredibly useful.
I just use list comprehension and lambda.

I set a condition in my Snapshot optimizer, turned on all pieces of gear and set the condition to >= Snapshot, so it would find every combination. It ran 544320 sets in 1.74 seconds. And my computer has 4G of RAM for reference, it's ***.
This is part of our misunderstanding. The scales of our problems are vastly different, so methods that work well for you will not necessarily work well in the situation I mentioned. If I consider Subtle Blow I/II, PDT, MDT, and DT, then I have 2.4 million times more sets to consider, which would take ~47 days to go through at 1.74 seconds per 540,000. You addressed this in your next comment, but I was convinced you were saying you went through 1 trillion sets in one second earlier...

Ramuh.Austar said: »
Ideally, you'd be able to turn off pieces of gear you have no use for or are redundant, you wouldn't need Rajas Ring on MNK to cap Subtle Blow, for example, so don't consider it.
You're right. Since the problem scales so quickly, removing options will quickly reduce the number of sets to check, which may allow the brute force method to work. I'm sure I could reduce this down to only have to check less than 1 billion. Further situational reduction would bring this down even more to a reasonable level.

The other part of our misunderstanding is that, even after finding however many sets that satisfy the given conditions, each set needs a few seconds to optimize for DPS on top of the conditional requirements on Subtle Blow, etc. This won't be a problem with only a few hundred sets to check, though.

Thanks, I'll look into this further. Reducing the amount of gear to check in each slot is definitely the best way to go if I wanted to use the brute force method over the iterative method.
 Siren.Noxzema
Offline
Serveur: Siren
Game: FFXI
user: Noxzema
Posts: 228
By Siren.Noxzema 2023-01-06 11:59:57
Link | Citer | R
  
So I ran Combo and it's showing to be the best ws for MNK lol. Though it isn't matching with what my Spreadsheet says dmg should be (26k, which is still not bad for Combo lol)

TP=[1500,2000] Min 44530 | Mean 91458 | Median 91218 | Max 124847

 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-06 12:03:46
Link | Citer | R
  
photobucket, really?
 Izanami
Offline
Posts: 298
By Izanami 2023-01-06 12:12:20
Link | Citer | R
  
Siren.Noxzema said: »
So I ran Combo and it's showing to be the best ws for MNK lol. Though it isn't matching with what my Spreadsheet says dmg should be (26k, which is still not bad for Combo lol)

TP=[1500,2000] Min 44530 | Mean 91458 | Median 91218 | Max 124847

This is partially because I made Combo a 5-hit weapon skill instead of a 3-hit weapon skill on accident.

The rest looks correct, but damage doesn't drop to 26k with just this correction. I can't view the image you linked because it requires an account, so I can't check details about this issue. Could you try imgur or something else that doesn't require an account?

I use the FTP values from Motenten's testing in 2007 2014 linked on BG Wiki since BG wiki does not provide FTP on the Combo page directly.
Combo parameters
 Siren.Noxzema
Offline
Serveur: Siren
Game: FFXI
user: Noxzema
Posts: 228
By Siren.Noxzema 2023-01-06 12:12:52
Link | Citer | R
  
Lol, I'm still stuck in 2015 I forgot all about Imgur, fixed it.
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-06 12:18:27
Link | Citer | R
  
Izanami said: »
wsc = 0.3*(player_str + player_dex) + dStat[1]*gearset.playerstats[dStat[0]]

it looks like you're not multiplying the gear stats by 0.3, just base stats here
 SimonSes
Offline
Posts: 8846
By SimonSes 2023-01-06 12:20:07
Link | Citer | R
  
Izanami said: »
Siren.Noxzema said: »
So I ran Combo and it's showing to be the best ws for MNK lol. Though it isn't matching with what my Spreadsheet says dmg should be (26k, which is still not bad for Combo lol)

TP=[1500,2000] Min 44530 | Mean 91458 | Median 91218 | Max 124847

This is partially because I made Combo a 5-hit weapon skill instead of a 3-hit weapon skill on accident.

The rest looks correct, but damage doesn't drop to 26k with just this correction. I can't view the image you linked because it requires an account, so I can't check details about this issue. Could you try imgur or something else that doesn't require an account?

I use the FTP values from Motenten's testing in 2007 linked on BG Wiki since BG wiki does not provide FTP on the Combo page directly.
Combo parameters

You have Combo parameters from before h2h update. It's now
1.00 2.40 3.40
[+]
 Siren.Noxzema
Offline
Serveur: Siren
Game: FFXI
user: Noxzema
Posts: 228
By Siren.Noxzema 2023-01-06 12:30:10
Link | Citer | R
  
With those 2 changes (3 hit and Ftp)
 Izanami
Offline
Posts: 298
By Izanami 2023-01-06 12:43:43
Link | Citer | R
  
SimonSes said: »
You have Combo parameters from before h2h update. It's now
1.00 2.40 3.40

Thanks. Is this the updated FTP list for H2H weapon skills? BG Wiki should really be updated for these FTP values. The MNK spreadsheet (09-2020) that I used for extra reference also shows the old Combo FTP values, despite being named in a way to suggest that it was updated after you found the new values.

BG claims Shijin is 1.5, but your 2019 post says 1.375. Should I also update my Shijin to 1.375? The values for the other weapon skills you tested back in 2019 seem roughly correct in my code.


Siren.Noxzema said: »
With those 2 changes (3 hit and Ftp)

I can't recreate this. I'm seeing half of that damage with nhits=3, and ftp=[1.0, 2.4, 3.4]. Your plot only shows 4 main peaks. These likely correspond to 5-hits, 6-hits, 7-hits, 8-hits. Are you sure you decreased nhits to 3? Your damage is also way higher for these peaks. There is something else going on in your code.

Either way, my damage is still way over 26k that your spreadsheet is seeing.
Combo updated FTP and 3-hit

Notice my plot shows 6 main peaks: 3-hits, 4-hits (DA), 5-hits (TA), 6-hits (TA+DA), 7-hits (TA+TA), 8-hits (TA+QA).


Ramuh.Austar said: »
Izanami said: »
wsc = 0.3*(player_str + player_dex) + dStat[1]*gearset.playerstats[dStat[0]]

it looks like you're not multiplying the gear stats by 0.3, just base stats here

At this point in the code, the player stats are the sum of base+gear+buffs, so that bit should be correct.
 SimonSes
Offline
Posts: 8846
By SimonSes 2023-01-06 12:53:05
Link | Citer | R
  
Izanami said: »
Thanks. Is this the updated FTP list for H2H weapon skills? BG Wiki should really be updated for these FTP values. The MNK spreadsheet (09-2020) that I used for extra reference also shows the old Combo FTP values.

BG claims Shijin is 1.5, but your 2019 post says 1.375. Should I also update my Shijin to 1.375? The values for the other weapon skills you tested back in 2019 seem roughly correct in my code.

This was first day tests and could have small errors. I would use http://wiki.ffo.jp/html/25607.html for highest accuracy probably. bg wiki copied most from JP wiki too, not sure why they left out things like Combo.

Izanami said: »
The values for the other weapon skills you tested back in 2019 seem roughly correct in my code.

If its sheet updated by me, then combo probably doesn't have "Updated" description on the right, which means its pre-updated values.
[+]
 Ramuh.Austar
Offline
Serveur: Ramuh
Game: FFXI
user: Austar
Posts: 10457
By Ramuh.Austar 2023-01-06 12:55:13
Link | Citer | R
  
Izanami said: »
BG claims Shijin is 1.5, but your 2019 post says 1.375. Should I also update my Shijin to 1.375? The values for the other weapon skills you tested back in 2019 seem roughly correct in my code.
BG is correct

at least for shijin, smite, the kicks, and raging since I did double check those for my simulations
[+]
 Siren.Noxzema
Offline
Serveur: Siren
Game: FFXI
user: Noxzema
Posts: 228
By Siren.Noxzema 2023-01-06 13:09:02
Link | Citer | R
  
Izanami said: »
Are you sure you decreased nhits to 3?

elif ws_name == "Combo":
base_ftp = [1.0, 2.4, 3.4]
ftp = np.interp(tp, base_tp, base_ftp)
ftp_rep = True
wsc = 0.3*(player_str + player_dex) + dStat[1]*gearset.playerstats[dStat[0]]
nhits = 3

 Izanami
Offline
Posts: 298
By Izanami 2023-01-06 13:18:28
Link | Citer | R
  
Siren.Noxzema said: »
Izanami said: »
Are you sure you decreased nhits to 3?
Spoiler Alert! Click to view.

Please triple check that you have one copy of the code and you've saved the weaponskill_scaling.py file after making the changes. I can recreate exactly what you're seeing only if I use nhits=5 and base_ftp = [1.0, 3.75, 5.5].
nhits=5, base_ftp=[1.0,3.75,5.5]
 Siren.Noxzema
Offline
Serveur: Siren
Game: FFXI
user: Noxzema
Posts: 228
By Siren.Noxzema 2023-01-06 13:31:12
Link | Citer | R
  
Yep, only 1 copy, and it's saved. Tried reloading and running it a few times, and still getting roughly the same.

Edit: Changed it back to the 1.0, 3.75, 5.5 and it's still the same.
Top Show Op
First Page 2 3 4 5 6 7 8 9
Log in to post.